One of the main challenges for computational intelligence is to understand how people detect and categorize objects in general, and process and recognize each other's face, in particular. The challenge becomes more daunting when one expands the biometric space to account for un-cooperative subjects, e.g., impostors, and has to handle temporal change, occlusion, and disguise. Towards that end we propose recognition-by-parts that integrates perception, learning, and decision-making using boosting and transduction. The architecture proposed facilitates layered categorization that starts with detection and proceeds with authentication. The strangeness / typicality concept supports both the representation and decision-making aspects responsible for pattern / target open set recognition. Typicality implements both the “filter” [feature selection and dimensionality reduction] and “wrapper” [classification] approaches. The parts, exemplar-based clusters of image patches, compete as weak learners during boosting using their typicality. The talk concludes with suggestions for augmenting and enhancing the scope and utility of the proposed architecture vis-à-vis active learning, multi-sensory integration and data fusion, change detection using martingale, and face selection and surveillance for CCTV.